Real-time positioning method for converter steel ladle vehicle

Abstract

The invention relates to a converter steel ladle vehicle real-time positioning method, which comprises the following steps of 1, buggy ladle data acquisition, 2, three-dimensional reconstruction of a buggy ladle, converting acquired three-dimensional point cloud data into a dense three-dimensional model of a converter steel ladle vehicle by using a dense three-dimensional reconstruction algorithm, and carrying out real-time positioning on the converter steel ladle vehicle. Thirdly, features of buggy ladle equipment are extracted; fourthly, the buggy ladle is positioned in real time. And 5, data processing and optimization. According to the technical scheme, real-time detection of the buggy ladle is realized in combination with the dense three-dimensional reconstruction model.

Description

Technical Field

[0001] This invention relates to a positioning method, specifically a real-time positioning method for converter ladle cars, belonging to the field of automation technology in iron and steel smelting. Background Technology

[0002] The ladle car in a converter is a crucial transportation device in the steel smelting process, and its precise positioning is essential for improving smelting efficiency and ensuring production safety. Traditional positioning methods rely on mechanical limit switches and laser rangefinders to control the ladle car's stopping position. After years of use, these two technologies have proven adequate for general ladle car parking. However, with the application of automated steel tapping technology in recent years, these two technologies are no longer sufficient. Mechanical limit switches can only control the position of a few fixed points and cannot achieve continuous vehicle position control. On some ground vehicles equipped with laser rangefinders, the rangefinder's measurement signal is transmitted to the control system via PROFIBUS communication or a direct 4-20mA analog signal. Due to the long travel distance of the ground vehicles and when the vehicles are in high dust levels below the furnace, the rangefinder's measured distance may suddenly change, or there may be signal instability such as a disconnection alarm. This makes it impossible to achieve precise position control of the vehicle during operation. Especially during automated steel tapping, incorrect ladle car position information can lead to abnormal adjustment of the ladle car's position and potentially cause steel spillage accidents. Therefore, developing various methods for real-time positioning of converter ladle cars is of great practical value and theoretical significance for improving the positioning accuracy and real-time performance of converter ladle cars.

[0003] Through retrieval and analysis of relevant literature, the following findings were obtained:

[0004] Patent 1, CN202210448161.1, describes a method for rapid positioning of a ladle car, which achieves positioning by marking the cable below the cable reel of the ladle car. This differs from this application.

[0005] Patent 2, CN202310339753.4, discloses a method and apparatus for locating the position of a converter ladle car. This method utilizes the state information to determine the vehicle's travel distance; constructs a coordinate system about the limit point with the length of the track as the abscissa; acquires the first position data measured by the position detection switch; uses the first position data as a verification value, and assigns the verification value each time the ladle car passes the position detection switch during its forward and reverse travel along the track, correcting the vehicle's travel distance; and determines the ladle car's position based on the corrected vehicle travel distance and its coordinate position on the abscissa. This method enables automatic and accurate positioning of the ladle car during the converter process. In summary, it can be seen that the technical problems solved by the existing disclosed solutions, the technical solutions adopted, and the effects achieved are all different from those of this invention. Therefore, a new solution is urgently needed to solve this technical problem. Summary of the Invention

[0006] This invention addresses the technical problems existing in the prior art by providing a method for real-time positioning of converter ladle cars. This technical solution combines a dense three-dimensional reconstruction model to achieve real-time detection of the ladle cars.

[0007] To achieve the above objectives, the present invention provides the following solution: a method for real-time positioning of a converter ladle car, the method comprising the following steps:

[0008] Step 1: Data collection from ladle cars.

[0009] Specifically as follows:

[0010] 1.1 Two RGB-D cameras are set up around the converter ladle car. Due to the relatively large amount of dust on site, cameras with high resolution are selected. The cameras are installed at the locations where images need to be captured, and parameters such as resolution and frame rate are set and the cameras are calibrated.

[0011] According to: P = K[R / T]P 1 Formula 1

[0012] In Formula 1: P is a point in the world coordinate system.

[0013] K is the camera intrinsic parameter matrix.

[0014] R is the rotation matrix.

[0015] T is the translation vector

[0016] P 1 Points in the camera coordinate system

[0017] 1.2 Collect 3D point cloud data of the surrounding area. This data should contain sufficient detail and texture information for subsequent 3D reconstruction, as follows:

[0018] 1.2.1 The camera acquires three-dimensional data, which is represented in the form of point cloud. Each point contains spatial coordinates (X, Y, Z) and color information RGB. The acquisition of the camera is triggered by the start of the ladle car. The start of the ladle car triggers the camera to work. The camera transmits the acquired images to the remote server through the network.

[0019] 1.2.2. Process the acquired images using 3D processing software and algorithms, including point cloud registration and surface reconstruction. The second step is the 3D reconstruction of the ladle car. Using a dense 3D reconstruction algorithm, the acquired 3D point cloud data is converted into a dense 3D model of the converter ladle car. This model should contain detailed surface information of the ladle car, such as shape and texture.

[0020] Specifically as follows:

[0021] 2.1 Perform inter-frame correspondence matching on the color and depth images received by the server in the first step, and perform feature point matching on the color and depth images of adjacent frames to establish inter-frame correspondence. The inter-frame correspondence matching method uses sparse feature points for preliminary matching, and then uses dense geometric and photometric information for detailed point cloud registration.

[0022] According to: E=\sum_{i=1}^{N}w_i||p_i-(Rp_i'+T) Formula 2

[0023] In formula 2:

[0024] E: Error

[0025] N: Number of points

[0026] w_i: Weight of each point

[0027] p_i: Source cloud

[0028] p_i': The corresponding point in the target point cloud

[0029] R: Rotation matrix

[0030] T: Translation vector

[0031] 2.2 Pose Estimation and Optimization: The camera pose is estimated based on the inter-frame correspondence and optimized to reduce errors.

[0032] 2.3 3D point cloud fusion: The 3D point cloud of each frame is fused into the global coordinate system to generate a dense 3D reconstruction model.

[0033] Third, feature extraction of the ladle car equipment, as detailed below:

[0034] Key feature points of the ladle car, including wheel positions and ladle edges, are extracted from the reconstructed 3D model. These feature points should possess good stability and identifiability for subsequent matching and localization.

[0035] Fourth, the real-time location of the ladle car is as follows:

[0036] By comparing 3D models at different time points, the displacement and attitude changes of the ladle car can be determined, enabling real-time localization. Feature point matching methods, such as the ICP (Iterative Closest Point) algorithm, can be used for model matching and localization.

[0037] Fifth, data processing and optimization, specifically as follows: Combining machine learning and deep learning technologies, preprocessing and feature extraction are performed on the collected image or point cloud data to improve the accuracy and efficiency of the reconstruction model. Simultaneously, filtering algorithms can be used to optimize the positioning data, further improving the accuracy and stability of the positioning.

[0038] Compared with existing technologies, this invention has the following advantages: 1. High positioning accuracy: Through dense 3D reconstruction and feature point matching, high-precision positioning of converter ladle cars can be achieved; 2. Strong real-time performance: Through continuous optimization of data processing and algorithm efficiency, real-time positioning of converter ladle cars can be achieved; 3. Strong adaptability: This method is applicable to different environments and usage scenarios, and has strong versatility and practicality; 4. It lays the foundation for automatic steel tapping technology in converters. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the process for organizing the present invention. Detailed Implementation

[0040] To enhance understanding of the present invention, the embodiments will be described in detail below with reference to the accompanying drawings.

[0041] Example: A method for real-time positioning of a converter ladle car, the method comprising the following steps:

[0042] Step 1: Data collection from ladle cars, as detailed below:

[0043] 1.1 Two RGB-D cameras are set up at the lifting position of the converter ladle car. The two cameras are placed at the two corners of the lifting position. Due to the relatively large amount of dust on site, high resolution cameras are selected. The cameras are installed at the positions where images need to be captured, and the camera resolution, frame rate and other parameters are set and calibrated.

[0044] 1.2 Collect 3D point cloud data of the surrounding area. This data should contain sufficient detail and texture information for subsequent 3D reconstruction, as follows:

[0045] 1.2.1 The camera acquires three-dimensional data, which is represented in the form of point cloud. Each point contains spatial coordinates (X, Y, Z) and color information RGB. The acquisition of the camera is triggered by the start of the ladle car. The start of the ladle car triggers the camera to work. The camera transmits the acquired images to the remote server through the network.

[0046] 1.2.2. Use 3D processing software and algorithms to process the acquired images, including point cloud registration and surface reconstruction.

[0047] The second step is the 3D reconstruction of the ladle car. A dense 3D reconstruction algorithm is used to convert the collected 3D point cloud data into a dense 3D model of the converter ladle car. This model should include detailed surface information of the ladle car, such as its shape and texture.

[0048] Specifically as follows:

[0049] 2.1 Perform inter-frame correspondence matching on the color and depth images received by the server in the first step, and perform feature point matching on the color and depth images of adjacent frames to establish inter-frame correspondence. The inter-frame correspondence matching method uses sparse feature points for preliminary matching, and then uses dense geometric and photometric information for detailed point cloud registration.

[0050] 2.2 Pose Estimation and Optimization: The camera pose is estimated based on the inter-frame correspondence and optimized to reduce errors.

[0051] 2.3 3D point cloud fusion: The 3D point cloud of each frame is fused into the global coordinate system to generate a dense 3D reconstruction model.

[0052] Third, feature extraction of the ladle car equipment, as detailed below:

[0053] Key feature points of the ladle car, including wheel positions and ladle edges, are extracted from the reconstructed 3D model. These feature points should possess good stability and identifiability for subsequent matching and localization.

[0054] Fourth, the real-time location of the ladle car is as follows:

[0055] By comparing 3D models at different time points, the displacement and attitude changes of the ladle car can be determined, enabling real-time localization. Feature point matching methods, such as the ICP (Iterative Closest Point) algorithm, can be used for model matching and localization.

[0056] Fifth, data processing and optimization, as detailed below:

[0057] By combining machine learning and deep learning technologies, preprocessing and feature extraction are performed on the acquired image or point cloud data to improve the accuracy and efficiency of the reconstruction model. Simultaneously, filtering algorithms can be used to optimize the localization data, further enhancing the accuracy and stability of the localization process.

[0058] Comparison of the implementation effects of this method during the experimental period:

[0059] This method accurately measures and locates the position of the ladle car in high-temperature, slag-rich furnace environments, replacing the traditional methods of stopping the ladle car by relying on mechanical impact gauges and single laser ranging. This method is technologically advanced, specifically in that: through a dense 3D reconstruction algorithm, the vehicle can be completely and accurately reconstructed, enabling real-time detection and analysis; based on dense point clouds and effective feature extraction, it effectively ensures the accuracy of relative position detection and status recognition. During the automatic tapping process in the converter, the ladle car's position moves with the converter's tilting angle. This precise and reliable positioning technology ensures accurate position feedback for the ladle car, effectively improving the alignment accuracy between the converter tapping spout and the ladle, eliminating iron spillage accidents caused by positional deviations, and playing a significant role in ensuring equipment safety and stability and reducing production accidents.

[0060] It should be noted that the above embodiments are not intended to limit the scope of protection of the present invention. Equivalent transformations or substitutions made based on the above technical solutions all fall within the scope of protection of the claims of the present invention.

Claims

1. A method for real-time positioning of a converter ladle car, characterized in that, The method includes the following steps: Step 1: Data collection from ladle cars. The second step is the 3D reconstruction of the ladle car. A dense 3D reconstruction algorithm is used to convert the collected 3D point cloud data into a dense 3D model of the converter ladle car. Third, feature extraction of ladle car equipment. Fourth, real-time positioning of the ladle car. Fifth, data processing and optimization.

2. The method for real-time positioning of converter ladle cars according to claim 1, characterized in that, Step 1: Data collection from ladle cars, as detailed below: 1.1 Two RGB-D cameras were installed around the converter ladle car. Due to the relatively high dust levels on site, high-resolution cameras were selected. The cameras were installed at the locations where images needed to be captured, and parameters such as resolution and frame rate were set and calibrated. According to: P = K[R / T]P 1 Formula 1, In Formula 1: P is a point in the world coordinate system. K is the camera intrinsic parameter matrix. R is the rotation matrix. T is the translation vector. P 1 Points in the camera coordinate system 1.2 Collect 3D point cloud data of the surrounding area. This data should contain sufficient detail and texture information for subsequent 3D reconstruction, as follows: 1.2.1 The camera acquires three-dimensional data, which is represented in the form of point cloud. Each point contains spatial coordinates (X, Y, Z) and color information RGB. The acquisition of the camera is triggered by the start of the ladle car. The start of the ladle car triggers the camera to work. The camera transmits the acquired images to the remote server through the network. 1.2.

2. Use 3D processing software and algorithms to process the acquired images, including point cloud registration and surface reconstruction.

3. The method for real-time positioning of converter ladle cars according to claim 2, characterized in that, The second step is the 3D reconstruction of the ladle car. A dense 3D reconstruction algorithm is used to convert the collected 3D point cloud data into a dense 3D model of the converter ladle car, as detailed below: 2.

1. Perform inter-frame correspondence matching on the color and depth images received by the server in the first step. Match feature points in adjacent color and depth images to establish inter-frame correspondences. The inter-frame correspondence matching method uses sparse feature points for initial matching, and then uses dense geometric and photometric information for detailed point cloud registration. According to: E=\sum_{i=1}^{N}w_i||p_i-(Rp_i'+T) Formula 2: In Formula 2: E: Error, N: The number of points, w_i: Weight of each point p_i: Source point cloud, p_i': The corresponding point in the target point cloud. R: rotation matrix, T: Translation vector 2.2 Pose Estimation and Optimization: The camera pose is estimated based on the inter-frame correspondence and optimized to reduce errors. 2.3 3D point cloud fusion: The 3D point cloud of each frame is fused into the global coordinate system to generate a dense 3D reconstruction model.

4. The method for real-time positioning of converter ladle cars according to claim 3, characterized in that, Third, feature extraction of the ladle car equipment, specifically as follows: extract the key feature points of the ladle car, such as the wheel positions and the edge of the ladle, from the reconstructed 3D model.

5. The method for real-time positioning of converter ladle cars according to claim 4, characterized in that, Fourth, the real-time positioning of the ladle car is as follows: by comparing the 3D models at different time points, the displacement and attitude changes of the ladle car are determined, and its real-time positioning is achieved. A feature point matching method, including the ICP (Iterative Closest Point) algorithm, is used to match and locate the model.

6. The method for real-time positioning of converter ladle cars according to claim 2, characterized in that, Fifth, data processing and optimization, as detailed below: By combining machine learning, deep learning and other technologies, the collected image or point cloud data is preprocessed and features are extracted to improve the accuracy and efficiency of the reconstruction model. At the same time, filtering algorithms are used to optimize the positioning data to improve the accuracy and stability of positioning.

Images